I don't even think it is a monkey's paw. That implies that it looked like it would solve a big problem in our initial estimation.

But "students will use the cheating machine to cheat" was obvious from the release of ChatGPT3. There was never some period of time where AI looked like it was a net positive for students only to be revealed to have an unexpected harm.

Even from the folks who claim to use LLMs to learn rather than cheat or avoid work, I've seen so many people admit that they are actually using it to harm themselves. "Oh, I only ask ChatGPT for the answer for really hard problems." Yeah man, doing the hard problems is how you learn.